As a personal learning project on how to write efficient GPU code I'm trying to implement different shortest-path algorithms on the GPU. This is notoriously one of the cases where a simple algorithm like Dijkstra on the CPU will beat many implementations on the GPU.

It has been a lot of fun to learn about Vulkan / GLSL and the GPU execution model to figure out why the CPU is so much faster than the GPU. I'll be open sourcing the code soon but so far I'm documenting my journey in a series of blog posts. First one of the series is https://www.execfoo.de/blog/deltastep.html